Fig. 1. Schematic of image acquisition

Fig. 2. Image acquisition process

Fig. 3. Background-suppression process and renderings of 3D-reconstructed image

Fig. 4. Principle of Graph cut

Fig. 5. Sample segmentation of Grab cut. (a) Original image; (b) segmentation process; (c) image after segmentation; (d) image after binary segmentation

Fig. 6. Framework of Grab cut

Fig. 7. Main improvement process from FCN to Deeplab v3

Fig. 8. Model-segmentation process of Deeplab

Fig. 9. Flow chart of algorithm combined Deeplab and Grab cut

Fig. 10. 3D-reconstruction effect of uncoated car. (a) Left-overlaid image; (b) right-overlaid image; (c) disparity map; (d) 3D point clouds

Fig. 11. Process of 3D-reconstruction image preprocessing using traditional method. (a) Original images; (b) background-suppression images; (c) segmentation images; (d) overlaid images

Fig. 12. Process of 3D-reconstruction image preprocessing using proposed method. (a) Original images; (b) background-suppression images; (c) segmentation images; (d) overlaid images

Fig. 13. Comparison of 3D-reconstruction effects and time consuming